Tokkibot

中文文档

Tokkibot is a general-purpose AI Agent that supports multi-channel interaction (CLI / Lark), tool invocation, long-term memory, and scheduled tasks.

Table of Contents

• ✨ Features
• 🚀 Quick Start
• 🛠 Usage
• 🔐 Environment Variables
• 📄 License

✨ Features

• Multi-channel Support: CLI interactive terminal, Lark (Feishu) group chat/IM bot
• Tool Invocation: File read/write, Shell execution, Web fetching, Skill extensions
• Subagent: Delegate complex tasks to specialized subagents
• Context Management: Auto-compression, history summarization, Token control
• Long-term Memory: Persistent memory across sessions
• Scheduled Tasks: Cron scheduling with result delivery to Lark
• Streaming Output: Real-time display of generated content

📅 Task Scheduling	🔍 Discovery	💻 Software Engineer	🧠 Learn & Memory

🚀 Quick Start

Initialization

tokkibot onboard

# Initialize workspace for a specific agent
tokkibot onboard --agent analyst

This creates the configuration file and workspace in ~/.tokkibot/.

Configuration

Edit ~/.tokkibot/config.json:

{
  "providers": {
    "openai": {
      "apiKey": "${OPENAI_API_KEY}",
      "baseURL": "https://api.openai.com/v1",
      "defaultModel": "gpt-4o-mini"
    },
    "deepseek": {
      "apiKey": "${DEEPSEEK_API_KEY}",
      "baseURL": "https://api.deepseek.com/v1",
      "defaultModel": "deepseek-reasoner",
      "enableThinking": true
    }
  },
  "agents": [
    {
      "name": "main",
      "maxIteration": 30,
      "provider": "openai",
      "model": "gpt-4o",
      "binding": {
        "match": { "channel": "lark", "account": "default" }
      }
    }
  ],
  "channels": [
    {
      "name": "lark",
      "account": {
        "default": { "appId": "your-app-id", "appSecret": "your-app-secret" }
      }
    }
  ]
}

🛠 Usage

CLI Interaction

# Start interactive TUI
tokkibot agent

# Single query
tokkibot agent --message "Write a Python script for me"

# Resume session
tokkibot agent --resume <session-id>

# Select specific agent defined in config
tokkibot agent --agent main

# Print current system prompt
tokkibot agent prompt

# List installed skills in CLI
tokkibot agent skills list

Gateway

tokkibot gateway

After starting, the gateway will listen to configured channels and route messages to the agent. Lark (Feishu) is one supported channel today, and more channels can be added later.

Supported Channels (Gateway only):

Channel	Status	Notes
Lark (Feishu)	✅	Group chat / IM bot integration

Control Commands:

Command	Description
/stop	Stop current task
/new	Start new session
/compact	Compress context
/skill list	List all available skills
/skill info <name>	Show skill details
/mcp list	List all MCP servers and status
/mcp info <server>	Show server tools
/model	Show current model and provider
/model set <provider> [model]	Switch provider/model
/status	Show current session status
/help	Show help

Scheduled Tasks

# List tasks
tokkibot cron list

# Add task
tokkibot cron add \
  --name "daily-report" \
  --expr "0 9 * * *" \
  --prompt "Generate daily work report"

# Add task with delivery
tokkibot cron add \
  --name "morning-greeting" \
  --expr "0 8 * * *" \
  --prompt "Generate a morning greeting" \
  --deliver \
  --channel lark \
  --to "oc_xxxxx"

# Add one-shot task (auto-disable after run)
tokkibot cron add \
  --name "deploy-check" \
  --expr "0 10 * * *" \
  --prompt "Run deployment checklist" \
  --once

# Manual execution
tokkibot cron run daily-report

# Enable/Disable
tokkibot cron enable daily-report
tokkibot cron disable daily-report

# Delete
tokkibot cron delete daily-report

Cron task definitions are stored in ~/.tokkibot/crons/. Task sessions use IDs like cron:<task-name>.

Skills

Skills extend the agent's capabilities with domain-specific knowledge and tools. Install skills using clawhub:

# Install a skill
clawhub install tavily-search --dir ~/.tokkibot/skills

# Install with specific version
clawhub install [email protected] --dir ~/.tokkibot/skills

Skills are automatically loaded from ~/.tokkibot/skills/ on startup. Each skill contains:

• SKILL.md - Skill definition and instructions
• Additional resources (prompts, templates, etc.)

MCP (Model Context Protocol)

Tokkibot supports MCP servers for extended tool capabilities.

You can configure MCP in two ways:

1. Use CLI (recommended):

# Add stdio MCP server to project config
tokkibot mcp add --transport stdio myserver -- npx -y @anthropic/mcp-server

# Add remote MCP server over HTTP
tokkibot mcp add --transport http remote-server https://api.example.com/mcp

# List and remove servers
tokkibot mcp list
tokkibot mcp remove remote-server

2. Edit config file manually (~/.tokkibot/mcp.json):

{
  "mcpServers": {
    "filesystem": {
      "command": "npx",
      "args": ["-y", "@anthropics/mcp-filesystem", "/path/to/allowed/dir"]
    },
    "github": {
      "command": "npx",
      "args": ["-y", "@anthropics/mcp-github"],
      "env": {
        "GITHUB_TOKEN": "${GITHUB_TOKEN}"
      }
    },
    "remote-server": {
      "url": "http://localhost:8080/mcp",
      "headers": {
        "Authorization": "Bearer ${API_TOKEN}"
      }
    }
  }
}

Configuration fields (command mode):

• command - Executable command to start the MCP server
• args - Command line arguments
• env - Environment variables (supports ${VAR} syntax for expansion)

Configuration fields (URL mode):

• url - Remote MCP server URL (streamable HTTP preferred, SSE fallback)
• headers - HTTP headers (supports ${VAR} syntax for expansion)

MCP config is loaded from both ~/.tokkibot/mcp.json and <project>/.tokkibot/mcp.json (project config overrides same-name global entries). MCP servers are started automatically and their tools become available to the agent.

🔐 Environment Variables

Variable	Description
OPENAI_API_KEY	OpenAI API Key
DEEPSEEK_API_KEY	DeepSeek API Key
MOONSHOT_API_KEY	Moonshot API Key

📄 License

Apache 2.0